Autoscaling Reconciliation with HPA

We should consider https://github.com/kedacore/keda rather than re-inventing the wheel, it keeps us a) relevant regarding the eco system, and b) constrains the model rather than letting management run wild with their demands!

In general the operator should generate events and conditions, then something should react to them.  The obvious ask is going to be for something monolithic, to which the answer is absolutely not.  By separating this into micro-services we can be a lot more agile with releases.  This is especially relevant when a customer starts using it, then decides on reflection it doesn't do what they want and needs a quick turn around before their service suffers due to scaling happening at totally the wrong time.